30 The reliability and construct validity of student perceptions of teaching quality C1. All average absolute difference scores were well below 0.1 (ranging from 0.00 to 0.03), indicating good model fit for the Impact! items. Second, the local independence of the items based on the association between two consecutive items was calculated. The results can be seen in Appendix C, Table C2. For all item pairs, the observed and expected values were approximately equal. In this investigation, the average absolute differences between the observed and expected values were also below 0.1 (ranging from 0.01 to 0.06), which is again an indication of good model fit for the Impact! items. In the third investigation, the observed and expected average scores for the two administration formats (paper-based and digital-based) were calculated (see Appendix C, Table C3 for the results). There was no differential item functioning between the two modes of administration, as the average absolute difference scores were all below 0.1 (ranging from 0.03 to 0.09). This also supports the good model fit of the Impact! items. All three investigations gave positive results regarding the model f it of the Impact! items. Considering this evaluation of the construct validity in a statistical manner, there is statistical support for the construct validity of the Impact! questionnaire. All items reflect one construct (hypothesized as teaching quality). 2.5.2 Reliability Global reliability The reliability coefficient for the estimation of teaching quality with an average of 19.07 students per class and 6.81 time points was .895. The parameter estimates of the model variances are provided in Table 2.2. The results of the g-study showed that 35.6% of the total variance was explained by differences between teachers, 24.4% of the total variance was explained by differences between students, and 12.7% by differences between time points. The interaction of different teachers and measurement moments explained 8.7% of the total variance. The proportion of unexplained variance was 18.7%.
RkJQdWJsaXNoZXIy MjY0ODMw